The business is releasing its next-generation version of the technology that drives the popular chatbot tool, ChatGPT, about four months after OpenAI startled the tech sector with it.
OpenAI announced GPT-4 in a blog post on Tuesday, claiming that it can score well on a variety of standardized tests and is also less likely to "go off the guardrails" with its responses, as some users have previously reported.
OpenAI said the revised system passed a simulated law school bar exam with a score around the top 10% of test takers; by comparison, the old version, GPT-3.5, scored around the bottom 10%. The business claims that GPT-4 can write code in all popular programming languages and read, analyze, or generate up to 25,000 words of text.
The update was referred to as the company's "latest milestone" by OpenAI. Even though it is still "less capable" than people in many situations, the corporation claims that it performs at a "human level" on a variety of professional and academic criteria.
The most recent iteration of OpenAI's big language model (GPT-4), which is trained on a considerable amount of web data, can produce engaging responses to user prompts. A few third-party applications, notably Microsoft's AI-powered Bing, are already incorporating the new version, which is currently only accessible via a waitlist.
“We are happy to confirm that the new Bing is running on GPT-4, which we’ve customized for search. If you’ve used the new Bing preview at any time in the last five weeks, you’ve already experienced an early version of this powerful model,” Microsoft stated on Tuesday.
Since its November 2022 launch, ChatGPT has won over many users with its capacity to produce unique essays, stories, and song lyrics in response to user questions, but it has also generated some controversy. In recent weeks, critics have criticized AI chatbots, notably those developed by Microsoft and Google, for being emotionally reactive, making factual mistakes, and outright participating in "hallucinations," as the industry describes it.
Similar restrictions apply to GPT-4 and previous GPT models. “It is still flawed, still limited, and it still seems more impressive on first use than it does after you spend more time with it,” the upgrade was announced in a series of tweets from OpenAI CEO Sam Altman on Tuesday.
However, he noted that there had been advances. He wrote that it is less prejudiced, hallucinates considerably less, and is more creative than earlier models. Yet, the business advised that "great care should be taken when using language model outputs, particularly in high-stakes contexts."
The information comes two weeks after OpenAI declared it was granting access to its ChatGPT tool to outside companies, opening the door for the chatbot to be included in a variety of programs and services.
Among the first partners exploring the service are Instacart, Snap, and the tutoring software Quizlet. After announcing in January that it was investing "multibillions of dollars" in OpenAI, Microsoft has now integrated the technology into a number of its products, including its Bing search engine.
Samantha Murphy Kelly, CNN Business, (2023 March 14th). "The technology behind ChatGPT is about to get even more powerful": Nearly four months after OpenAI stunned the tech industry with ChatGPT, the company is releasing its next-generation version of the technology that powers the viral chatbot tool.